brew/Library/Homebrew/formula_installer.rb
Max Howell fd3ddf8a72 Disable watch_out_for_spill temporarily
It's a great feature but it seems to conflict with our new system() replacement. Clearly the bug is still in our system() but what exactly it is is not clear and this bug sucks.

See Homebrew/homebrew#124
2009-11-11 17:29:34 +00:00

73 lines
1.8 KiB
Ruby

require 'beer_events'
require 'formula'
require 'set'
class FormulaInstaller
@@attempted = Set.new
def initialize
@install_deps = true
end
attr_writer :install_deps
def expand_deps f
deps = []
f.deps.collect do |dep|
dep = Formula.factory dep
deps += expand_deps dep
deps << dep
end
deps
end
def install f
expand_deps(f).each do |dep|
begin
install_private dep unless dep.installed?
rescue
#TODO continue if this is an optional dep
raise
end
end if @install_deps
install_private f
end
private
def install_private f
return if @@attempted.include? f.name
@@attempted << f.name
# 1. formulae can modify ENV, so we must ensure that each
# installation has a pristine ENV when it starts, forking now is
# the easiest way to do this
# 2. formulae have access to __END__ the only way to allow this is
# to make the formula script the executed script
read, write = IO.pipe
# I'm guessing this is not a good way to do this, but I'm no UNIX guru
ENV['HOMEBREW_ERROR_PIPE'] = write.to_i.to_s
begin #watch_out_for_spill do #disabled temporarily, see Issue #124
fork do
begin
read.close
exec '/usr/bin/ruby', '-I', File.dirname(__FILE__), '-rinstall', f.path, '--', *ARGV.options
rescue => e
Marshal.dump(e, write)
write.close
exit! 1
end
end
ignore_interrupts do # because child proc will get it and marshall it back
write.close
Process.wait
data = read.read
raise Marshal.load(data) unless data.nil? or data.empty?
raise "Suspicious installation failure" unless $?.success?
end
end
end
end